T. Kawaguchi1, L. Yan2, Q. Qi2, S. Liu2, K. Takabe1 1Roswell Park Cancer Institute,Breast Surgery, Department Of Surgical Oncology,,Buffalo, NY, USA 2Roswell Park Cancer Institute,Department Of Biostatistics & Bioinformatics,Buffalo, NY, USA
Introduction:
MicroRNAs (miRNAs) are small non-coding RNAs that exert its functions by regulating expression of their target genes. Dysregulations of miRNAs are related with breast cancer (BrCa).The purpose of this study was to classify BrCa with miRNAs expression patterns to predict survival utilizing The Cancer Genome Atlas (TCGA).
Methods:
Both clinical and miRNA-seq data enrolled in TCGA dataset were retrieved from the GDC data portal for analyses, and were evaluate by hierarchical clustering based on bioinformatics analysis. We also evaluate clinical relevance including prognostic analysis based on the novel subclasses using the Cox proportional hazard model.
Results:
Of 1097 cases enrolled in TCGA dataset, 1052 caseswere used for miRNAs expression data and survival analysis. We devided the cases into “short” (died within 3 years after diagnosis), “long” (lived longer than 5 years), and the others. We identified that 15 miRNAs (let-7a-1, miR-106a, miR-17, miR-184, miR-18a, miR-193a, miR-19a, miR-20a, miR-20b, miR-362, miR-4661, miR-500a, miR-766, miR-92a-1, miR-93) were significantly differently expressed between the long and short group. With the expression pattern of these 15 miRNA, the patients were classified into three clusters. Of the 15 miRNAs, we conducted additional feature selection in a multivariate Cox proportional hazard model, and three miRNAs remain after model selection (miR-106a, miR-766, and miR-93). We generated a risk scoring model with the expression of the three miRNAs based on Cox proportional hazard model. We found that the patients with the high score significantly associated with poor outcome.
Conclusion:
We demonstrated a novel concept that microRNA expression patterns of BrCa can predict worse survival.